3876
T. Ji et al. / Bioorg. Med. Chem. Lett. 14 (2004) 3875–3879
DNA double helix. However, the two proteins also bind
to each other, and therebystabilize the ternarycomplex,
where HOXB1 places a short peptide arm (FDWMK)
into a hydrophobic binding pocket (a ‘hot spot’) on the
surface of PBX1. Previous in vitro studies have shown
that disruption of the cooperative DNA binding by
HOX to PBX1 proteins can be accomplished bypoint
mutations in this HOX peptide domain.14 These studies
also showed that the Trp 182 (W) and Met 183 (M)
residues within the highlyconserved pentapeptide region
of the HOX familyof peptide domains (F/Y-P/D- W-M-
K/R) were critical for cooperative DNA binding with
PBX. The crystal structure shows HOXB1 Trp 182 and
Met 183 packing against each other and binding in the
hydrophobic pocket on the surface of PBX1.13
This cooperative binding can be competitivelyblocked
Figure 1. As shown, the naphthalene ring is positioned
within the hydrophobic pocket, the C-1 amide side chain
NH is forming a hydrogen bond with Leu-252 and the 3-
fluoro atom on the phenyl amide side chain is hydrogen
bonding with Try-291. The C-4 ether amide side chain is
extending across the surface of PBX1 in the direction of
the bound DNA. From these modeling studies the C-4
side chain appeared to have multiple potential binding
opportunities so a range of side chains at this position
was selected for experimental evaluation (Table 1).
A focused combinatorial libraryof thirt-ytwo 1,4-
disubstituted naphthalene derivatives with the modeled
3-fluorophenyl amide C-1 side chain and various ether
amide side chains at C-4 were synthesized (Fig. 1). The
synthesis begins from commercially available 4-meth-
oxy-1-naphthaldehyde 1 as outlined in Figure 1. Oxi-
dation of aldehyde 1 provided acid 2.17 Amide 3 was
obtained bycoupling the corresponding acid chloride
with 3-fluoroaniline. Demethylation of methyl ether 3
using AlCl3/EtSH produced the corresponding phenol,18
which was then converted to methyl ester 4. Acid 5 was
obtained byrefluxing ester 4 with LiOH in methanol.
The final target amides 6 were synthesized using PyBOP
as the coupling reagent with a broad range of amines.
All final products were purified bysilica gel chroma-
tographyto greater than 95% purityand gave 1H NMR
and MS spectra that are consistent with the expected
product.
bythe 12-residue peptide, QPQI YPWMRKLH (IC50
100 lM), containing the conserved pentapeptide
¼
sequence.14
The human HOXB1–PBX1/DNA (code 1B72) complex
described above was downloaded from the Protein Data
docking, and librarydesign studies were carried out
using SYBYL 6.8, and the associated modules FlexX,
CScore, and LeapFrog, all obtained from Tripos, Inc.
that binds in the hydrophobic pocket on the surface of
PBX1 as described above was deleted to expose the
critical peptide-binding region for designing small mol-
ecule antagonists, and the resulting complex was mini-
The selectivityof individual members of the libraryfor
the HOXA1–PBX1/DNA target, as well as the abilityof
some analogs to cross-over to other transcription factor
targets was evaluated bytesting them in parallel against
the BRN1 and BRN2 transcription factors. BRN1 and
BRN2 are members of mammalian class I POU tran-
scription factor familyand are expressed in the devel-
oping embryonic brain.19 POU domain class
transcription factors contain, in addition to the POU
domain, a homeodomain helix-turn-helix class DNA
binding motif similar to that found in the HOX and
PBX class transcription factors.20 EMSAs were per-
formed with BRN1 and BRN2 exactlyas was done with
HOXA1–PBX1 except that a sequence recognition ele-
ment known to bind BRN1/BRN2 was used as the
probe.21
mized.
A
varietyof potential scaffolds, with
representative appended side chains, were docked into
the now exposed hydrophobic pocket and the sur-
rounding surface of PBX1. These candidate scaffolds
were selected bya combination of visual evaluation of
the binding surface, hand docking of candidate ligands,
automated docking (FlexX), de novo design experiments
(LeapFrog), ease of synthesis, and predictions of binding
affinity(FlexX and CScore). An important criteria that
was also applied to the candidate scaffold selection pro-
cess is the abilityto produce potential antagonists with
MW<500 and that have the abilityto meet the additional
15
‘rule of 5’ criteria developed byLipinski et al.
for
compounds likelyto be successful as oral therapeutics.
Finally, rigid scaffolds that result in antagonists with a
limited number of rotatable bonds were given priority
since this rigidityis also predicted to improve the prob-
Of the 32 analogs listed in Table 1, 24 showed measur-
able inhibition of the formation of the HOX1–PBX1/
DNA ternarycomplex when screened at a 300 lM initial
concentration. A relativelyhigh initial screening con-
centration was chosen for two reasons: (1) the known
12-residue peptide antagonist, QPQIYPWMRKLH,
containing the conserved pentapeptide sequence
(underlined) that binds to the Pbx1 hydrophobic pocket,
has an IC50 of only100 lM, (2) protein–protein inter-
action targets are significantlymore challenging to block
with small molecules than the classical drug targets for
which lower screening concentrations can be used.
16
abilityof obtaining orallyactive drugs.
With the above criteria in mind the first scaffold selected
for synthesis and testing was the 1,4-disubstituted
naphthalene scaffold (Table 1). This rigid scaffold pro-
vides two diversityside chains able to interact with the
PBX protein at the mouth of the hydrophobic pocket
and a phenyl ring to penetrate into the hydrophobic
pocket. The libraryof potential antagonists prepared
from this scaffold all have molecular weights between
340 (7) and 489 Da (36) and have a limited number of
freelyrotating bonds (Table 1).
Of these 24 active compounds against the HOXA1–
PBX1/DNA target the most potent were 30 and 31 with
an IC50’s ¼ 86 and 65 lM, respectively. These com-
A modeled complex (after minimization) of the parent
inhibitor based upon this scaffold, 7, is illustrated in